Seeing the big picture

Bill Buxton, chief scientist for Toronto’s Alias I Wavefront is fighting an uphill battle. It’s not easy getting developers to recognize that products in the technology sector need to be designed rather than engineered. And when it comes to data visualization tools, design can make all the difference.

Several years ago Australia, officials at a pressure petro (propane) chemical factory were starting to worry about their equipment – it was getting old and replacement parts were scarce. The company decided to buy a computer system to help monitor reactor vessels, which would hopefully help extend the life of the equipment through timely warnings.

They found a system with beautiful 3D graphics and implemented it. The workers, who did not receive proper training on the system, were at a loss. More importantly, no one had thought to ask them how they got the job done before the system arrived. It turned out they relied more on physical cues as to when knob A needed to be turned or lever B should be pulled, according to Gitte Lindgaard, director of human-oriented technology at Carleton University in Ottawa. Some of the cues would have included steam coming out of number three, or a funny smell, she said.

“The assumption behind the design of the new system was that the people using it would be at a high cognitive level, taking complex data into consideration to make decisions. This was not the case,” said Lindgaard, who worked for the petro factory at the time.

The factory was only two kilometres away from airports and a city with a population of four million. As Lindgaard said, “one big bang and you could probably say goodbye to half of those people. That’s not a good place to do things wrong.

“The assumptions were greater than the investigation of the people.”

The psychology behind data visualization systems is paramount. Buxton noted that humans do not change in terms of how technology runs. There is no Moore’s Law for humans. Only after developers understand people can they be opportunistic about technology.

Largely, people are visually stimulated, and Buxton said it would be hard to argue that complex phenomenon can’t be simplified by rendering them in visual form. However he said, there has to be an understanding about relationships or data visualization will buy you nothing.

Gareth Morfill, a senior analyst with London, England-based Ovum, said data visualization developers need to understand cognitive aspects to build useful tools.

Often data visualization will represent more than six variables in one graphical element. The hope is that users, by looking at that picture, will gain insight they would not have if they were only looking at numbers.

One example of data visualization at its finest is a map of Napoleon’s 1812 march into Russia, by French engineer Charles Minard. On the map there are longitudinal and latitudinal lines, along with place names and a band that depicts Napoleon’s path (with retreats in different colours). The thickness of the band represented the number of soldiers Napoleon had marching with him. On the bottom is a timeline with corresponding temperatures. Without words, Minard tells the history of this failed march.

Morfill said that by just looking at that map it was so easy to see the variables and the impact of different moves. “Data visualization and understanding our ability to cognitively understand it, while packing in as much information as possible into a single view, is so much more complicated than throwing up a bar chart,” he said.

Lindgaard said it really wouldn’t matter how much information is thrown at the person if the developer could not anticipate users’ questions.

“The information doesn’t actually take any shape inside someone’s head unless it addresses something that they want to have addressed,” Lindgaard said.

If a company tried to take information from a PC and give it visually to a PDA user, they would have to ask themselves what the person on the run needs to know. What questions will they be asking? They may not need all the information, and simply minimizing all the charts to PDA size isn’t going to cut it, Lindgaard said.

“Given that few vendors are psychologists and the psychological literature is not terribly accessible, we consequently have many information systems and other types of applications that really do a poor job for the operator,” she said.

looking for takers

Regardless of whether they are doing everything they are meant to do, data visualization tools are on the front burner for many companies.

It seems to be one of the areas in technology that is bubbling, Morfill said.

Michael MacDonald, CEO of Rockville, Md.-based Visual Mining, a data visualization solutions provider, said he is seeing data visualization a lot more, likely because it is becoming easier to do. “You can’t manage what you can’t measure, unless you can see the information, and data visualization is the best way to do that,” he said.

With the number of financial scandals in the past year, knowledge management and data management have become incredibly important. Falling in line with quick data retrieval and easy-to-spot patterns and anomalies, is data visualization.

Matching time and project management with campaign goals and log in times, can show who is giving their most to which campaign and whether or not work can be done to make it better.

IMC Corp. in Bell Plains, Sask., recently started using Visual Mining’s NetCharts Server 4.0. Brian Crosthwaite, manager of application development for IMC, said they were mainly concerned with plant applications for the Bell Plains potash mine.

“Our data could have hundreds of thousands of points,” Crosthwaite said. “To look at it in an Excel spreadsheet could take hours to decipher. With data visualization we can see dips and anomalies in a graph quickly.”

Joel Gaab, system analyst for IMC, said most of what the company is doing with their system is driven by what operations people are demanding to see.

Don Campbell, vice-president of products for Ottawa-based Cognos Corp., is sure that data visualization tools will reach a high acceptance point. “If you put a lot of information in a pie chart, it doesn’t always make sense, but if you put it on a map, it will make more sense.”

Tim Bray, CTO and founder of Vancouver-based Antarctica Systems Inc., which sells a visual browser engine, said he has seen more interest in data visualization lately. He pointed to the success of business intelligence (BI) vendors such as Cognos and Crystal Decisions in these tough economic times, saying one of the important things they bring to the table is visual access to a certain class of enterprise data.

“Even in boring, back-end fields like enterprise storage, there’s a feeling that the management level of visibility into what they’re getting isn’t good enough,” Bray said.

Powerful PCs can generate powerful graphics, and Bray said that now nearly everyone has a heavy-duty PC on their desk and this has led to more interest in being able to do these visualizations.

Buxton noted that falling hardware prices generally mean falling software prices, and data visualization tools are no exception – another reason why there are more examples of enterprises having access to these tools.

looking for acceptance

But Morfill is unsure whether the interest in data visualization will amount to any real acceptance of the technology. “It’s been around for a long time,” he said. “It’s been talked about and thought of as having great applications, but it’s rarely implemented – like artificial intelligence.

“We know it would be good for us, but we can’t find the application.”

Despite increasing interest, visualization tools still have a fairly high learning curve, Morfill said, hindering them from being ubiquitous.

“Much like data mining, there has been a promise that we will get all these useful nuggets of information and all this insight. Yet, it has not gained mass market appeal – data visualization is like that,” Morfill said.

Although MacDonald might argue there are many early adopters of data visualization tools, and that it is moving into the mainstream, he does note that a lot of people are quite happy with basic pie charts and spreadsheets.

“There are some pretty nifty tools out there for doing things,” he said. “You can plot five different points. They don’t necessarily use the same scale, but it allows you at a glance – you plot a point on each axis – and it gives you a picture of what is happening.”

One of the things Antarctica hit its head on was 3D. Bray said the company had been sure it was the way to go. “It should work. There are millions playing video games everyday, and you’d think there was a place for 3D in business data, but boy, it doesn’t seem to take flight.”

A lack of visualization is more likely to hit the senior manager than the employee on the floor, who is gaining good information from reports. The senior manager doesn’t want to be worried about inventory counts on hand; they need a bigger picture of the enterprise as a whole.

“When you’re trying to take a really big picture combining a lot of sources, it’s hard to see how you can do that really effectively at all without a visual interface,” Bray said.

Buxton said data visualization might be suffering from the next-big-thing syndrome. “Basically, the market for PCs and standard applications is saturated. So everybody’s looking for the next big thing.”

He cautioned that “data visualization” is quickly becoming a buzzword, which would make it as meaningless as “user friendly.”